Computing and Informatics

Refined descriptive sampling is designed to improve upon the descriptive sampling method for experimentation in simulation. The former reduces significantly the risk of sampling bias generated by descriptive sampling and eliminates its problem related to the sample size. In this paper, we propose an optimal parallel Monte Carlo simulation algorithm using refined descriptive sampling and evaluate in parallel architecture, performance measures of a stable M/M/1 queueing system, a Pert network and the Newsboy problem.